Structural insights into the role of SHOC2-MRAS-PP1C complex in RAF activation

Abstract

RAF activation is a key step for signaling through the mitogen-activated protein kinase pathway. The SHOC2 protein, along with MRAS and PP1C, forms a high affinity, heterotrimeric holoenzyme that activates RAF kinases by dephosphorylating a specific phosphoserine. Recently, our research, along with that of three other teams, has uncovered valuable structural and functional insights into the SHOC2-MRAS-PP1C (SMP) holoenzyme complex. In this structural snapshot, we review SMP complex assembly, the dependency on the bound-nucleotide state of MRAS, the substitution of MRAS by the canonical RAS proteins, and the roles of SHOC2 and MRAS on PP1C activity and specificity. Furthermore, we discuss the effect of several RASopathy mutations identified within the SMP complex and explore potential therapeutic approaches for targeting the SMP complex in RAS/RAF-driven cancers and RASopathies.

Graphical Abstract

The process of RAS-mediated RAF activation involves the dephosphorylation of a specific phosphoserine in RAF located between the cysteine-rich and kinase domains. This crucial step is accomplished by the SHOC2-MRAS-PP1C complex. Recently, multiple research groups have determined the structures of this complex, which reveal how the complex assembles, its role in Noonan syndrome, how it specifically dephosphorylates RAF, and possible ways for therapeutic interventions.

Introduction

The mitogen-activated protein kinase (MAPK) signaling pathway amplifies extracellular signals through RAS-mediated activation of the RAF, MEK, and ERK protein kinases. These signals regulate cell survival, growth, differentiation and proliferation [1]. The RAS-mediated activation of RAF is a key step in the MAPK signaling pathway. RAF kinases (ARAF/BRAF/CRAF) are held in an autoinhibited, monomeric state by the highly conserved, regulatory phosphoserine/phosphothreonine-binding 14-3-3 proteins in the cytosol [2, 3]. RAF kinases contain multiple phosphoserine residues [4]. Two of these phosphoserines that flank the kinase domain of RAF interact with a dimer of 14-3-3. One of them is located within Conserved Region 2 (CR2) and hereby referred to as CR2-pS (ARAF S214, BRAF S365, CRAF/RAF1 S259), while the second is found in Conserved Region 3 (CR3) and hereby referred to as CR3-pS (ARAF S582, BRAF S729, CRAF S621) [5]. The RAF activation cycle starts when active RAS interacts with the RAS-binding domain (RBD) in the autoinhibited RAF complex bringing RAF to the plasma membrane [6–8]. Although the exact order of subsequent steps is not known, several essential steps must occur for RAF activation; (i) the cysteine-rich domain (CRD) must disengage from both the RAF kinase domain and 14-3-3 dimer in the autoinhibited monomeric complex, and interact with both RAS and the plasma membrane; (ii) the kinase domain needs to undergo rearrangement and form a dimer; (iii) a 14-3-3 dimer holding the CR2-pS and CR3-pS of RAF in cis needs to reorganize and stabilize dimeric RAF by binding to the CR3-pS of two RAF monomers. To prevent 14-3-3 engagement with the CR2-pS site which restricts RAF to an autoinhibited conformation, specific dephosphorylation of the CR2-pS must occur during the RAF activation process [7]. This dephosphorylation event is carried out by the SHOC2-MRAS-PP1C holoenzyme (SMP) complex and is essential for the full activation of RAF and MAPK signaling [9, 10].

SHOC2-MRAS-PP1C (SMP) Complex

Among the three proteins that form the SMP complex, SHOC2 is the least understood. Based on the sequence analysis, SHOC2 is predicted to be primarily comprised of leucine-rich repeats (LLRs), with the first 90 residues being intrinsically disordered [11]. LRR proteins are involved in many different cellular functions, including innate immunity, cell adhesion, receptor kinases, apoptosis, autophagy, mRNA transport and neuronal development [12]. Multiple repeats of the highly variable 22-30 amino acid LRR domain allow the formation of horseshoe- or arc-shaped proteins which have a low sequence identity, can carry out diverse functions but are structurally similar [13, 14]. Several LRR-containing proteins, including SCRIB and ERBIN, are known to bind and regulate SHOC2 function in the context of the SMP complex [15–17]. Despite a limited understanding of SHOC2, it has repeatedly been shown to be a therapeutic target in oncogenic RAS-driven cells. SHOC2 was shown in multiple studies to be the strongest synthetic lethal target in the presence of MEK inhibitors in KRAS mutant lung and pancreatic cancer cell lines [18–20]. Furthermore, the Broad Institute’s Cancer Dependency Map (DepMap) [21], shows a strong dependency of oncogenic G12, G13 and Q61 mutants of HRAS, KRAS and NRAS on SHOC2 for cancer cell growth and survival [22]. MRAS shares ~50% sequence identity with the canonical RAS proteins and are regulated by the same GEFs and GAPs as H/K/NRAS [23, 24]. However, unlike canonical RAS proteins, MRAS is a poor activator of the MAPK pathway due to a lower affinity for RAF arising from residue differences in its interswitch region [9, 24, 25].

In the SMP complex, the dephosphorylation activity is carried out by the protein phosphatase 1 catalytic domain (PP1C). PP1C is a manganese- or iron/zinc-dependent metalloenzyme [26]. The two active site metals are located at the center of three converging substrate channels named Hydrophobic, Acidic or C-terminal channel [27]. Substrates bind to one or two of these channels to place the phospho-serine or -threonine into the active site [28, 29]. In humans, PP1C exists as three highly conserved (>90% sequence identity) isoforms, PP1CA, PP1CB and PP1CC [30]. PP1C alone has low activity and specificity, but gains activity through an unknown mechanism when bound to PP1C interacting proteins (PIPs) [31]. PIPs typically use short linear motifs (SLiMs) to engage with PP1C. The RVxF motif is the most understood of these SLiMs, as it is found in >200 different PIPs [32]. PIPs can redistribute PP1C to specific cellular locations, aid in substrate recruitment, and alter the specificity of PP1C by blocking or extending one of the active site channels [32–35]. Multiple germline mutations in SHOC2, MRAS, and PP1CB proteins have been detected in Noonan syndrome (NS) and Noonan-like syndrome with anagen hair (NSLH or Mazzanti syndrome), types of RASopathy disorders (Fig. 1A) [36–46]. To explore the structural and functional role of SHOC2 and MRAS in the recruitment of PP1C to the plasma membrane and dephosphorylation of the CR2-pS in RAF kinases, four research groups, including us, recently determined the structure of the SMP complex by cryo-EM and X-ray crystallography [22, 47–49]. These studies also determined the affinity of SMP assembly by biophysical methods, its dependence on GTP-bound MRAS, and that MRAS could be substituted for the canonical H/K/NRAS. Both structural and biophysical data provided a rationale for why certain NS and NSLH mutations cause aberrant MAPK signaling. Altogether, these structural studies of the SMP complex improve our understanding of SMP assembly, why certain RASopathy mutations in the SMP complex cause dysregulation of MAPK signaling, and the recognition of RAF substrates by the SMP complex.

Fig. 1. Structure of the SHOC2-MRAS-PP1C complex.

(A) Domain organization of SHOC2, MRAS and PP1C. All NS and NSLH mutations currently identified in the SMP complex are highlighted and mutations present at a protein-protein interface are shown in bold. PP1C NS mutations are denoted using PP1CB numbering. (B) Cartoon representation of the SHOC2-MRAS-PP1CA complex shown in two orientations (PDB 7TVF) is colored green, blue and pink, respectively. Switch I and II of MRAS are shown in light blue and purple, respectively, and the active site manganese ions in PP1C are shown as gray spheres. This figure was generated using PyMOL [72].

Assembly and structure of the SMP complex

Active SMP complex only forms in the presence of GTP-bound MRAS. Both isothermal titration calorimetry (ITC) and surface plasmon resonance (SPR) experiments showed that all three proteins assemble into a ternary complex with an affinity of 30-350 nM (depending on the experimental technique) [22, 47–49]. Interestingly, no stable binary interaction was observed for any of the proteins, except possibly a weak transient interaction between SHOC2 and PP1C by all four groups, suggesting that this forms first [22, 47–49]. This is supported by the observations that increasing concentrations of SHOC2 (compared to MRAS) enhances PP1C catalytic activity towards RAF substrates [22], and the SMP complex only formed when SHOC2 is exposed to PP1C followed by MRAS in a bio-layer interferometry experiment [48]. Hauseman et al. observed a weak complex between SHOC2 and MRAS, while Liau et al. observed complex formation between GDP-bound MRAS with all isoforms of PP1C [22, 49]. The interaction of PP1C with GDP-bound MRAS may allow PP1C to remain at the plasma membrane, waiting for activation through both MRAS nucleotide exchange and SHOC2 binding.

The structure of the SMP complex independently determined by four different research groups shows a similar arrangement of three proteins. We co-expressed SHOC2 and the pre-assembled SMP complex using the LRR folding chaperone, SUGT1, to enhance their expression [47, 50], while the other three groups purified individual proteins and assembled the complex before purifying it again [22, 48, 49]. Two structures, by Liau et al. and Kwon et al. were determined by cryo-EM to a resolution of ~2.9 Å, while Hauseman et al. and Bonsor et al. used X-ray crystallography to a resolution of 1.95 Å and 2.17 Å, respectively [22, 47–49, 51, 52]. Each X-ray structure contained two copies of the SMP complex in the asymmetric unit. Despite the slight differences in constructs, mutations, PP1C isoform and structural determination method used, all SMP complexes superimpose with each other with low RMSDs (a range of 0.4-1.8Å). All structures were determined in the presence of either GTP or a non-hydrolyzable GTP analogue. SHOC2, the only component of the SMP complex whose isolated structure was unknown, was also solved by all four groups. As predicted, SHOC2 was found to contain 20 LRR domains, which adopt a horseshoe-like shape. The overall structure of the complex indicates that SHOC2 acts as a platform for the assembly and stability of the SMP complex, bringing together the other two components, MRAS and PP1C, to synergistically form the functional holoenzyme (Fig. 1B). Both MRAS and PP1C interact extensively with the concave surface of SHOC2. MRAS uses both Switch-I and Switch-II as well as a loop in the allosteric lobe to engage with 14 of the 20 LRR domains. PP1CA, like MRAS, engages with most LRR domains, but mainly via SHOC2 residues located on the periphery of the concave molecular surface. SHOC2 also uses a non-canonical RVxF SLiM (the arginine of the RVxF motif in SHOC2 is a glycine – G63) in its intrinsically disordered N-terminal domain to form additional contacts with PP1C via a β-hairpin (Fig. 1B). The RVxF motif is missing in the SHOC2 construct used by Hausemen et al., though their complex is stabilized by the introduction of three NS mutations in a LRR of SHOC2 (M173I), MRAS (Q71R) and PP1CA (P50R). The hydrophobic valine and phenylalanine of the SHOC2 RVxF SLiM are buried in a hydrophobic pocket on PP1C in an identical manner to other RVxF-containing PIPs. Kwon et al. performed a systematic deep mutagenesis scan of every residue within the SHOC2 protein [48]. Their scan corroborated that residues of the LRRs of SHOC2, which interact with MRAS or PP1C, are essential for SMP complex formation. Furthermore, mutating the SHOC2 RVxF motif to a canonical RVxF motif score as a gain of function mutation. Specifically, the SHOC2-G63R was found to increase the binding of MRAS and PP1C [48]. MRAS and PP1C interactions with SHOC2 place them adjacent to each other where they form additional contacts. MRAS uses its N-terminal, pre-switch I and interswitch residues to bind PP1C (Fig. 1B). PP1CA-R188 is the only residue in the entire SMP complex that contacts the other two proteins. Our mutagenesis study showed that this residue is essential for forming the SMP complex [47]. All residues of SHOC2 and MRAS are at least ~20Å away from the active site, with all three active site channels fully solvent-exposed (Fig. 1B).

Role of NS and NSLH mutations in SMP complex formation

Recurrent mutations in SHOC2, MRAS or PP1C have not been detected in cancer except SHOC2-T411S, which is observed in liver cancer [53]. However, multiple mutations have been detected within SHOC2, PP1CB and MRAS in NS and NSLH patients (Fig. 1A). The MRAS NS mutations, G23V, T68I and Q71R have been found to keep MRAS in the constitutively active, GTP-bound state [40, 41]. The Q71R mutation also forms van der Waal interactions with SHOC2 as observed in the Hauseman et al. structure [49]. The most common NS mutation in SHOC2 is the S2G mutation, which constitutively targets all cellular SHOC2 to the plasma membrane through the creation of a de novo N-myristoylation site [46]. However, several other NS mutations, such as G53R, M173I/V, Q269R, T411A, and two double mutations M173I_L174F and Q269H_H270Y are observed in SHOC2 [36, 37, 45]. Currently, all seven NS mutations observed in PP1C are only seen in the PP1CB isoform of PP1C [42–44, 54–56]. Two of these seven, P49R, and E183A (PP1CB numbering), are present at the PP1C-SHOC2 interface. The structure of the SMP complex reveals that the NS mutations observed in the SMP complex interfaces either increase van der Waal contacts (SHOC2-M173I/V, SHOC2- M173I_L174F or SHOC2-Q269H_H270Y), form a de novo hydrogen bond (SHOC2-Q269R and PP1CB-P49R), or relieve charge-charge repulsion (PP1CB-E183A) or steric hindrance (SHOC2-T411A). Interestingly, the methyl group of SHOC2-T411 was noted by Kwon et al. to be destabilizing towards SMP complex formation, before its recent identification in two NS patients where it was found to be mutated (T411A) to relieve this steric hindrance [36, 37, 48]. The role of the NS mutation, SHOC2-G53R, remains unclear as this region of SHOC2 is in the intrinsically disordered region which is not resolved in any of the structures. However, this region has been shown to be involved in the nuclear export of SHOC2 to the cytoplasm. The SHOC2-G53R may enhance its nuclear export and affect the cellular distribution of SHOC2 [57]. Several of these mutants (SHOC2-M173I, SHOC2-Q269H_H270Y, MRAS-Q71R, PP1CB-P49R and PP1CB-E183A) were observed to form a ternary complex with 2-4-fold higher affinity by SPR [22, 47–49]. The SHOC2-M173I, PP1CB-P49R and PP1CB-E183A mutants all display a slower off-rate compared to the wild-type complex [47]. In vivo experiments have shown that several SHOC2 NS mutants (S2G, G53R, M173I, Q269R, T411A) increase the concentration of active SMP complexes resulting in sustained levels of phospho-ERK [36]. Taken together, these data suggest that NS mutations extend the lifetime of the SMP complex, causing sustained dephosphorylation of RAF and prolonging MAPK signaling.

SHOC2 and MRAS activation of PP1C

In the SMP complex, all three proteins play an essential role in carrying out the dephosphorylation of RAF substrates. PP1C alone has poor catalytic activity and specificity as a phosphatase towards RAF and other substrates in general [32, 58, 59]. The SMP holoenzyme is unusual compared to other PP1C holoenzymes as it requires two PIPs to stimulate catalytic activity. Liau et al. showed that individually SHOC2 or MRAS (to a lesser extent) could weakly stimulate PP1C catalytic activity, but together, synergistically lead to significantly enhanced catalytic activity of PP1C [22, 47]. Furthermore, MRAS plays an additional role as it causes SMP complex formation to occur at the plasma membrane. This plays an important role in colocalizing the SMP complex near the RAS-RAF complex, whereas PP1C carries out the dephosphorylation reaction using its active site channels (Fig. 2A). The exact mechanism by which PIP binding increases the catalytic activity of PP1C is not fully understood. As no structural changes of PP1C are typically observed upon complexation with PIPs, this suggests the catalytic activity of PP1C is not allosterically regulated by PIP binding to the RVxF pocket. However, the NS mutation in PP1CB-A56P is found adjacent to the RVxF binding pocket. We observed that the mutation has no effect on SMP assembly though it causes aberrant signaling. We venture that PP1CB-A56P could be catalytically active independent of SHOC2 and MRAS binding [47].

Fig. 2. Recognition of RAF substrates by the SMP complex.

(A) Electrostatic surface of PP1C (red – acidic, white – neutral and blue – basic) with the three active site channels shown as dashed lines. The polybasic linker (RKK) of SHOC2 between the RVxF motif and LRRs is shown as a green dashed line. (B) Sequence alignment centered around the phosphoserine of the CR2-pS and CR3-pS sites of RAF kinases. Conserved residues are highlighted in black. The acidic residues of the RAF CR3-pS +1 site are highlighted in gray. Proposed C-terminal hydrophobic residues of the CR2-pS, which bind to the hydrophobic groove of SHOC2, are shown in beige. Sequence alignment was generated using Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/) and presented using ESPript 3 (http://espript.ibcp.fr/ESPript/ESPript/). UniProt accession numbers for ARAF, BRAF and CRAF are P10398, P15056 and P04049, respectively. (C) Docked model of BRAF CR2-pS peptide (beige ribbon) onto the SMP complex. The peptide occupies the acidic and hydrophobic channel of PP1C. We hypothesize that the C-terminal hydrophobic residues of CR2-pS bind to the hydrophobic groove of SHOC2. The C-terminal half of SHOC2 shows high flexibility, shown in green and gray, between the two copies of the SMP complex found in the asymmetric unit (PDB 7TVF). This domain movement may aid in the removal of the CR2-pS from 14-3-3. Figures shown in panels A and C were generated using PyMOL [72].

Certain PIPs display PP1C isoform specificity [60]. However, no changes in SMP complex formation nor catalytic efficiency were observed through the substitution of the three isoforms of PP1C in the SMP complex [22, 47]. PIP complexes where PP1C isoform specificity has been demonstrated typically contact regions of PP1C which show sequence differences between the isoforms. From the structures, we observe that SHOC2 nor MRAS contact these residues [22, 47–49].

Dephosphorylation of RAF substrates

The SMP complex specifically dephosphorylates CR2-pS of RAF substrates. Other phosphorylation sites such as CR3-pS, pS445 (BRAF) and pS338 (RAF1) are not dephosphorylated by the SMP complex [9, 15, 47, 50]. Although PP1C alone has poor activity towards CR2-pS, high concentrations of PP1C can dephosphorylate CR2-pS [47]. This suggests that the specificity of PP1C towards RAF CR2-pS substrates is already present and that MRAS and SHOC2 increase the activity of PP1C towards RAF [22, 47]. The CR2-pS +1 site in RAF substrates is either alanine or threonine, while the +1 site at the CR3-pS or other phosphoserine sites in RAF are glutamic or aspartic acids (Fig. 2B). PP1C has previously been noted to prefer small residues and disfavor acidic residues at the +1 position [61]. Four NS mutations have been identified in or around the active site channels of PP1CB; H124R, R220C, D252Y and E274K [54–56]. As these mutations are more than ~20 Å away from any of the protein-protein interfaces, it is possible that these NS mutations affect RAF substrate binding or catalytic activity of PP1C rather than SMP complex assembly.

We performed docking of 15-mer RAF CR2-pS peptides to the SMP complex, which revealed that most of the peptides were placed with their N and C termini occupying the acidic and hydrophobic channels (Fig. 2C) [47]. Liau et al. identified that the residues C-terminally of the CR2-pS are conserved and several were hydrophobic (Fig. 2B) [22]. Furthermore, they identified that the C terminus of SHOC2 contains a hydrophobic groove that is located adjacent to the hydrophobic channel of PP1C. Molecular dynamic simulations were used with RAF CR2- and CR3-pS to observe which formed stable complexes. Only RAF CR2-pS complexes were stable, interacting with both the hydrophobic channel and groove of PP1C and SHOC2, respectively [22]. Interestingly, the structure of the SMP complex determined by us, showed that five residues of the C-terminal disordered region of PP1C interact with the hydrophobic groove of the C terminus of SHOC2 identified by Liau et al. suggesting a similar mode of binding by CR2-pS (Fig. 2B) [22, 47]. We speculate that the C terminus of PP1C may inhibit RAF binding to the hydrophobic groove of SHOC2. Furthermore, this autoinhibitory mechanism could be regulated through phosphorylation of T320 of PP1C, which would prevent PP1C interaction with the hydrophobic groove of SHOC2 [62].

SHOC2 and MRAS binding to PP1C occlude most of the known PIP binding sites on PP1C, thereby preventing the formation of other PIP-PP1C complexes at the plasma membrane and the dephosphorylation of their substrates [47]. Certain PP1C-PIP complexes can act as non-competitive inhibitors toward specific substrates. Nuclear inhibitor of PP1 (NIPP1) and myosin phosphatase target 1 (MYPT1) were initially identified as inhibitors of PP1C, as these complexes prevented dephosphorylation of the canonical substrate, glycogen phosphorylase a [63, 64]. However, it was soon realized that these “inhibitors” prevented the dephosphorylation of a range of substrates yet remained active against certain substrates [65, 66]. The structures of NIPP1-PP1C and MYPT1-PP1C complexes revealed that NIPP1 and MYPT1 use two different SLiMs linked by an intrinsically disordered polybasic region to interact with PP1C [60, 66]. The polybasic region in the MYPT1-PP1C complex was found to surround the entrance to the active site acidic channel of PP1C (the polybasic region of NIPP1 is unresolved but is in the vicinity of the acidic channel entrance). Mutation of the polybasic region of NIPP1 and MYPT1 restores the specificity of PP1C towards glycogen phosphorylase a [66]. We note that SHOC2 binds to PP1C through two SLiMs; the RVxF and LRR binding domains. These two SLiMs are linked by a small disordered polybasic linker (residues 77-86). Structures of the SMP complex show the polybasic region, like MYPT1 and NIPP1, would be located at the entrance of the acidic channel (Fig. 2A). This suggests that the SMP complex could act as a non-competitive inhibitor towards other substrates, preventing specific substrates from being dephosphorylated and competing with RAF dephosphorylation at the plasma membrane by the SMP complex.

One important mechanistic part that remains unclear is the potential competition between 14-3-3 and SMP complex for the CR2-pS. RAF is held in an autoinhibited state through the CR2-pS binding to 14-3-3. For dephosphorylation to occur, the CR2-pS needs to be released from the 14-3-3. Mechanistically, the SMP complex must either play an active role in the displacement of 14-3-3 or allow the dissociation of CR2-pS from the 14-3-3 passively to carry out dephosphorylation.

The SMP complex could actively bind to the solvent-exposed hydrophobic residues present downstream of the CR2-pS site by interacting with the C-terminal hydrophobic groove of SHOC2 (Fig. 2B-C) [22]. The two SMP complexes in the asymmetric unit determined by us showed differences in the C-terminal half of SHOC2 [47]. In one of the SMP complexes, the last 10 LRRs are displaced away from MRAS and PP1C with the loss of several important interactions (Fig. 2C). This hinge movement of SHOC2 was predicted before the structure of SHOC2 was determined [57]. We speculate that this SHOC2 hinge movement could aid in the displacement of 14-3-3 bound to the CR2-pS through tugging and weakening the affinity of the CR2-pS towards 14-3-3 (Fig. 2C). This would allow dephosphorylation of the CR2-pS, prevent its reassociation with 14-3-3 and cause rearrangement of the 14-3-3 population, with now a 14-3-3 dimer aiding in the formation of RAF dimers through only engaging with the CR3-pS.

The alternative mechanism could involve SMP complex binding to the CR2-pS once it dissociates from the 14-3-3. Although bivalent binding of 14-3-3 to RAF CR2-pS and CR3-pS is likely to be of high affinity (as observed in other 14-3-3-binding systems [67]), the off-rate of 14-3-3 to the CR2-pS may be high enough to allow the SMP complex dephosphorylating phospho-serine without actively displacing 14-3-3.

MRAS vs canonical RAS isoforms

As MRAS shares significant sequence similarity with the canonical RAS proteins, GTP-bound HRAS, KRAS and NRAS were tested for their ability to form a complex with SHOC2 and PP1C [22, 47–49, 68]. It was found that active SHP, SKP and SNP complexes could form, though these complexes assemble with a K_D ~10-20-fold weaker affinity when compared to SMP complex formation [22, 47–49]. From the SMP complex structures, we observe that three different regions of MRAS could explain why MRAS preferentially forms higher affinity interaction with SHOC2 and PP1C over H/K/NRAS proteins. Two regions arise at the MRAS-PP1C interface (Fig. 3A). MRAS contains an additional ten amino acids at its N terminus which are not present in H/K/NRAS (Fig. 3B). These residues form a hydrogen bond and van der Waal interactions with PP1C (Fig. 3A). Deletion of the N-terminal residues of MRAS results in an 8-fold weakening of SMP complex formation [47]. The other difference is found in the interswitch region (Fig. 3A). MRAS shows several residue differences in this region compared to H/K/NRAS (Fig. 3B). Young et al. has previously shown these interswitch region differences are important for SMP complex assembly [9]. The final difference centers around MRAS-H132, which forms a hydrogen bond to SHOC2-E428 (Fig. 3C). The corresponding region in H/K/NRAS is one residue shorter and would not be able to form this contact (Fig. 3D). Mutation of this region in MRAS to the residues observed in HRAS or KRAS results in a 3-4-fold weakening of SMP complex formation [47].

Fig. 3. SHOC2 and PP1C preferentially bind to MRAS over the canonical RAS proteins.

(A) MRAS-PP1C interface showing interactions formed by MRAS residues present in the N-terminal extension and interswitch region. These residues are either absent or substituted by different amino acids in H/K/NRAS proteins. (B) Sequence alignment of the N-terminal extension and interswitch region of MRAS, KRAS, HRAS and NRAS. (C) MRAS-SHOC2 interface showing MRAS-H132 located at the C-terminal region forming a hydrogen bond to SHOC2. Figures shown in panels A and B were generated using PyMOL [72]. (D) Sequence alignment of the C-terminal region of MRAS, KRAS, HRAS and NRAS. Totally conserved residues are highlighted in black and key differences observed in MRAS are boxed in yellow in panels (B) and (D). Sequence alignments were generated using Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/) and presented using ESPript 3 (http://espript.ibcp.fr/ESPript/ESPript/). UniProt accession numbers for HRAS, KRAS, MRAS and NRAS are P01112, P01116, O14807 and P01111, respectively.

Currently, no structure of SKP, SHP or SNP has yet been determined. Liau et al. demonstrated that SKP could be deposited on grids for cryo-EM, but 2D class averages revealed a strong orientation preference [22]. Optimization efforts to prevent this preference resulted in the dissociation of the SKP complex. As H/K/NRAS form weaker complexes compared to MRAS due to the differences described above, it may be possible to preserve the SHP/SKP/SNP complexes through mutations that stabilize complex assembly (Fig. 3B, D) [9, 47].

Therapeutic targeting of the SMP complex

In normal situations, MRAS would be the preferred binding partner of SHOC2 and PP1C, due to the 10-20-fold higher affinity over the canonical RAS proteins as described above. This is a reversal of what is observed with RAF binding where the canonical RAS proteins preferentially bind to RAF over MRAS [25]. In vivo studies have supported the fact that the lower affinity SHOC2-H/K/NRAS-PP1C and MRAS-RAF complexes are able to function in some situations [69–71]. We hypothesize that in an oncogenic RAS environment, the constitutively active GTP-bound RAS mutant proteins (G12X, G13X and Q61X) are not only localizing RAF to the plasma membrane but also forming SHOC2-RAS_oncogenic-PP1C complexes, driving dephosphorylation of the CR2-pS site concurrently. Small molecules targeting oncogenic RAS mutants in the GDP- or GTP-bound states that alter the conformation of switch regions, a strategy employed to target RAS-RAF interactions, would not only prevent binding to RAF but also inhibit SHOC2-RAS_oncogenic-PP1C complex formation and therefore dephosphorylation of RAF CR2-pS.

Binding studies suggest no stable binary complex forms between any two of the three proteins. Therefore, targeting any of the binary interfaces should prevent active ternary complex formation. Disruption of the PP1C-SHOC2 or PP1C–MRAS interfaces may be difficult to achieve as over 200 PIPs are known to interact with PP1C [32, 34]. Most of these sites are occluded by SHOC2 or MRAS, therefore any small molecule targeting the SHOC2 or MRAS binding sites on PP1C would most likely cause off-target effects [47]. However, PP1CA-R188, an essential residue involved in the assembly of the SMP complex, is found not to occupy any of the currently known PIP binding sites on PP1C [47]. Targeting this residue would potentially prevent SMP assembly without disrupting other PP1C-PIP complexes. A small molecule binding to SHOC2 would have to target the SHOC2-MRAS interface as the PP1C interface is rather narrow and shallow. Switch I and II binding surface of SHOC2 has been suggested as a possible druggable site, specifically LRRs 2-5. LRR 3 is comprised of small residues, while LRR 2 and 4 are larger, creating a groove where a possible drug could bind and prevent RAS binding [47].

Unresolved questions

Despite independent studies by multiple groups on the SMP complex, several questions remain unanswered. Further studies are required to; (1) understand if 14-3-3 plays any role in the regulation of RAF dephosphorylation by the SMP complex; (2) identify whether oncogenic RAS proteins, in addition to recruiting RAF to the plasma membrane, drive dephosphorylation through SHOC2-PP1C complexes in cancer cells; and (3) provide a deeper understanding of how the SMP complex interacts and dephosphorylates the RAF–RAS complex at the plasma membrane. These further studies would enable novel therapeutic approaches for targeting RAS/RAF-driven cancers and NS.

Abbreviations

CR2

Conserved Region 2

CR2-pS

Conserved Region 2 phosphoserine

CR3

Conserved Region 3

CR3-pS

Conserved Region 3 phosphoserine

CRD

Cysteine-rich domain

GEFs

Guanine nucleotide exchange factors

H/K/NRAS

HRAS, KRAS and NRAS

K_D

Dissociation constant

LLR

Leucine-rich repeat

MAPK

Mitogen-activated protein kinase

MYPT1

Myosin phosphatase target 1

NIPP1

Nuclear inhibitor of PP1

NS

Noonan syndrome

NSLH

Noonan-like syndrome with anagen hair

PIP

PP1C interacting protein

PP1C

Protein phosphatase 1 catalytic domain

RBD

RAS binding domain

SHP

SHOC2-HRAS-PP1C

SKP

SHOC2-KRAS-PP1C

SLiM

Short linear motif

SMP

SHOC2-MRAS-PP1C

SNP

SHOC2-NRAS-PP1C